CS Spring 2010 CS 414 – Multimedia Systems Design Lecture 32 – Multimedia OS Klara Nahrstedt Spring 2010

Administrative MP4 is out, Start early Competition on May 3, 2010, 5-6:30pm in 216 SC Deadline of MP4 for grading – Saturday, May 1, 3-5pm in 216 SC (pre-competition to decide on the finalists) All should come, pizza and 1,2,3 rd prices of the competition will be provided between 6:30-7pm on May 3 CS Spring 2010

CPU Process/Thread Scheduling Maintain many programs in memory Determine how to best schedule them Requires information about the processes and an understanding of the system goals Switch among processes rapidly so each user perceives direct ownership of system CS Spring 2010

CPU Scheduling Evaluation Criteria Throughput: processes completed per unit time Turnaround: from submission to termination Wait: time waiting in the ready queue Response: from user request to system response Utilization: how busy the CPU is over time Want predictable system behavior CS Spring 2010

Scheduling Algorithms FCFS Shortest Job First Priority scheduling Round-robin Multi-level queue Rate monotonic Earliest deadline first For General Purpose Process CPU Scheduling For RT/Multimedia Process CPU Scheduling CS Spring 2010

Real-time/Multimedia Processing Requirements Need to process continuous multimedia data  Processing occurs in predetermined, usually periodic intervals  Processing must be completed by certain deadlines Need RT process manager  Perform admission control  Determine schedule  Perform reservation  Schedule to give processing guarantees CS Spring 2010

RT/Multimedia Processing Requirements Main Problem: How to find a feasible schedule? Conflicting Goals/Problems: How do we schedule multimedia (RT) processes so that 1. non-RT processes do not starve when RT process is running 2. RT process is not subject to priority inversion CS Spring 2010

Model in RT Scheduling Task (Process) – schedulable unit Task characterized by  Timing constraints  Resource requirements Assumptions  Periodic tasks without precedence relations Time constraints  s – task starting point  e – task processing time  d – task deadline  p – task period CS Spring 2010

Periodic Constant Processing Time Tasks Example Usage Pattern: Peak Processing Time=10ms Period = 100ms Time(ms) CS Spring 2010

Periodic Variable Processing Time Tasks Example Usage Pattern: Period = 100ms Peak Processing Time = 30ms Sustainable Processing Time = 15ms Burst Tolerance = 7ms Time(ms) CS Spring 2010

Model of RT Scheduling Congestion avoidance deadline  If period at (k-1) step is equal to ready (start) time of period k Tasks: preemptive vs. non-preemptive Main goal of RT Scheduling:  find feasible schedule of all periodic tasks so that newly arriving task and all previous admitted tasks finish processing in every period according to their deadline CS Spring 2010

Model of RT Scheduling Must have Schedulability (Admission) Test for RT tasks What is the performance metric for RT tasks?  Guarantee ratio := number of guaranteed tasks/total number of tasks  Process utilization (U): CS Spring 2010

Scheduling Policies of RT Tasks Rate- Monotonic Scheduling (RMS)  Designed/proved by C.L. Liu and Layland 1973  Policy: task with highest rate has highest priority  Static and optimal, priority-driven Optimal means that there no other static algorithm that is able to schedule a RT task which can’t be scheduled by RMS algorithm Assumptions:  Tasks are periodic  Each task must complete before next request  All tasks are independent  Run-time of each task request is constant  Any non-periodic task has no required deadline CS Spring 2010

Example of RMS CS Spring 2010

Scheduling Policies for RT Tasks Earliest Deadline First (EDF) Policy  Optimal dynamic algorithm  Produces valid schedule if one exists  Complexity O(n 2 )  Upper bound of process utilization 100%  Policy: task with earliest deadline has highest priority CS Spring 2010

Example of EDF CS Spring 2010

Comparison between RMS and EDF CS Spring 2010

Admission Control (for preemptive tasks) Schedulability test for RMS Schedulability test for EDF CS Spring 2010

Example Consider the following preemptive RT tasks and their characteristics  T1: p1 = 50ms, e1=10ms  T2: p2 = 100ms, e2=20ms  T3: p3 = 200ms, e3=50ms  T4: p4 = 100ms, e4=20ms Are these tasks schedulable via RMS?  If yes, what is the feasible schedule? Are these tasks schedulable via EDF?  If yes, what is the feasible schedule? CS Spring 2010

Example of Multimedia Process Management System (DSRT) - Execution Flow of a SRT Process Probe Admission Control Monitor/ Conformance Test Probing Phase Reservation Phase Execution Phase Adaptation Extract Reservation Contract Processor Binding Adjusted Contract SRT Process Scheduling CS Spring 2010

Multiprocessor Partitioning Design Guaranteed PartNon-guaranteed Part PCPT Processes PVPT Processes RT SchedulerOverrun SchedulerTS Scheduler TS Processes Reserved RunOverrun BurstReserved RunOverrun Processor #1 Processor #2 RT PartitionOverrun PartitionTS Partition RT PartitionOverrun PartitionTS Partition CS Spring 2010

Partition Scheduling RT SchedulerOverrun SchedulerTS Scheduler Processor #1 Processor #2 RT PartitionOverrun PartitionTS Partition RT PartitionOverrun PartitionTS Partition Top-Level Scheduler Preemptive EDF Multi-Level Round Robin Priority Queues Kernel Scheduler Proportional Sharing CS Spring 2010

RT Partition Scheduler Waiting Queue: (Sorted with the earliest released time) Processor #1 p p p Runnable Queue: (Sorted with the earliest deadline) (1) admitted (2a) finished one iteration (2b) overrunning (3) released for next iteration Overrun Partition Queues CS Spring 2010

Overrun Partition Scheduler Burst Queue (FIFO) Overrun Queue (FIFO) Permanent Non-conforming Queue (FIFO) p (1a) Conformed? (1b) Nonconformed? (1c) Nonconformed frequently? p (2) Miss Deadline Lowest Priority Highest Priority CS Spring 2010

How to Implement SRT in Best Effort Linux (User-level Priority Dispatch) DSRT Scheduler Process Dispatched Process (3) Scheduler sets Timer for T ms. (1) Increase Priority. (2) Set Processor Affinity. (5) Decrease Priority Highest Priority Lowest Priority Time Sharing Processes RT Process Pool Fixed RT Priority Dynamic TS Priority (4) Scheduler sleeps. (5) Dispatched Process runs for T ms. (6) Timer interrupts Scheduler N N CS Spring 2010

Experimental Result SRT1: MPEG player With P=100ms SRT2: MPEG player with P=50ms. CS Spring 2010

Conclusion RMS and EDF are basic policies for real- time scheduling systems For multimedia systems, soft-real-time scheduling (SRT) concepts needed, connecting reservation-based and adaption-based SRT DSRT system shows implementation of multimedia SRT systems CS Spring 2010